Estimation: How to Catch a Wrong Answer Before You Submit It

Most wrong answers in mental math are not wrong by a little; they are wrong by a lot, and a two-second estimate would have shown it. Estimation is the skill of knowing roughly what an answer should look like before, during and after you calculate it. It is the cheapest accuracy upgrade there is.

Why estimate at all?

A calculation has two ways to fail: a small slip (a digit off by one) or a structural error (a misplaced decimal point, a missed carry in the hundreds, an operation done in the wrong order). Estimation will not catch small slips, but it reliably catches the structural ones, and those are the errors that cost the most. If a quick estimate says "about 400" and your answer is 4,000, you know immediately to look again.

Technique 1: Round to one significant digit

Round each number to a single leading digit, calculate with those, and treat the result as a target. For 683 × 41, use 700 × 40 = 28,000. The exact answer, 28,003, matches almost perfectly. For 4,860 ÷ 19, use 5,000 ÷ 20 = 250 (the exact answer is 255.8).

Example: is 37 × 52 = 1,824 plausible?

Round to 40 × 50 = 2,000. 1,824 is close, so the answer is plausible. The exact answer is 1,924, so 1,824 is off by 100, and the rough check was too coarse to see it; see technique 3.

When rounding, a sensible rule is to round one number up and the other down in a product so that the errors cancel out. In 37 × 52, rounding 37 up to 40 and 52 down to 50 gives a closer answer than rounding both up.

Technique 2: Know which direction your estimate errs

A useful estimate is not only "about" the answer; it also tells you whether the exact value is larger or smaller:

  • If you round both factors of a product up, the true product is smaller than your estimate.
  • If you round the numerator up and the denominator down, the true quotient is smaller than your estimate.
  • If you round the number you subtract up, the true difference is larger.

Knowing the direction lets you bracket the answer. For 48 × 52, 50 × 50 = 2,500 is a rough middle, but 48 × 52 = 50² − 2² = 2,496 sits just below it.

Technique 3: Check the last digit and the digit sum

Rough size checks catch big errors; two cheap exact checks catch small ones.

The last digit

The last digit of a sum or product depends only on the last digits of the numbers you started with. In 37 × 52, 7 × 2 = 14, so the answer must end in 4. The wrong answer 1,824 also ends in 4, so this check does not help here. It does flag a wrong answer such as 1,925 immediately.

Casting out nines

Add the digits of each number repeatedly until you reach a single digit (this is the number's remainder when divided by 9, with 9 itself treated as 0). Apply the same operation to those single digits; the result must agree with the digit sum of your answer.

Check 37 × 52 = 1,924

37 → 3 + 7 = 10 → 1. 52 → 5 + 2 = 7. Multiply the check digits: 1 × 7 = 7.
Answer: 1,924 → 1 + 9 + 2 + 4 = 16 → 7. They match ✓.
Now test the wrong answer 1,824: 1 + 8 + 2 + 4 = 15 → 6. That does not equal 7, so 1,824 is wrong ✓.

This test can only prove an answer wrong, never absolutely right (two errors can cancel, and a swap of digits is invisible to it), but it is remarkably good at catching a wrong digit.

Technique 4: Use benchmarks you already know

Compare against landmark numbers: 1⁄2, 1, 10, 100. Is 3⁄7 more or less than a half? (Less, because half of 7 is 3.5.) Is 0.46 × 38 more or less than 19? (Slightly less, since 0.46 is a bit under 0.5.) Benchmarking turns an unfamiliar calculation into a comparison with something easy.

Technique 5: Count the digits

The number of digits in an answer is predictable. A two-digit times a two-digit number gives a three- or four-digit result. A four-digit number divided by a two-digit number gives a two- or three-digit result. Before you type an answer, ask whether it has the right number of digits. Decimal-point mistakes disappear almost completely once you build this habit.

Make estimating automatic

  1. Before you calculate, say a rough range out loud: "somewhere between 1,500 and 2,000".
  2. Calculate exactly.
  3. Compare. If the answer falls outside the range, stop and redo the calculation before you submit.

In timed races this feels like it costs speed, but it actually saves it: a wrong answer in Competitive costs you time to retry, and a quick sanity check is cheaper than a retry.

Race a Competitive match → More guides